Local Decorrelation For Improved Pedestrian Detection
Authors: Woonhyun Nam, Piotr Dollar, Joon Hee Han
NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate boosted decision tree learning with decorrelated features in the context of pedestrian detection. As our baseline we utilize the aggregated channel features (ACF) detector [7], a popular, top-performing detector for which source code is available online. Coupled with use of deeper trees and a denser sampling of the data, the improvement obtained using our locally decorrelated channel features (LDCF) is substantial. While in the past year the use of deep learning [25], motion features [27], and multi-resolution models [36] has brought down log-average miss rate (MR) to under 40% on the Caltech Pedestrian Dataset [10], LDCF reduces MR to under 25%. This translates to a nearly tenfold reduction in false positives over the (very recent) state-of-the-art. |
| Researcher Affiliation | Collaboration | Woonhyun Nam Strad Vision, Inc. woonhyun.nam@stradvision.com Piotr Doll ar Microsoft Research pdollar@microsoft.com Joon Hee Han POSTECH, Republic of Korea joonhan@postech.ac.kr |
| Pseudocode | No | The paper does not contain any pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | No | The paper states, 'To further improve clarity, all source code for LDCF will be released.' and 'While the proposed locally decorrelated channel features (LDCF) require only modest modification to existing code, we will release all source code used in this work to ease reproducibility.' These are future promises, not current availability. |
| Open Datasets | Yes | Current practice is to use the INRIA Pedestrian Dataset [6] for parameter tuning, with the test set serving as a validation set, see e.g. [20, 2, 9]. We utilize this dataset in much the same way and report full results on the more challenging Caltech Pedestrian Dataset [10]. |
| Dataset Splits | Yes | Of the 71 minute long training videos ( 128k images), we use every fourth video as validation data and the rest for training. On the validation set, LDCF outperforms ACF by a considerable margin, reducing MR from 46.2% to 41.7%. |
| Hardware Specification | No | The paper mentions 'timed using 12 cores' but does not specify any particular CPU model, GPU, or other hardware details like memory or specific cloud instances. |
| Software Dependencies | No | The paper does not provide specific version numbers for any software dependencies or libraries used. |
| Experiment Setup | Yes | For these results we used regularization of ϵ = .1 and patch size of m = 5... For all remaining experiments we fix k = 4 and m = 5... We first augment model capacity by increasing the number of trees twofold (to 4096) and the sampled negatives fivefold (to 50k). |